In [1]:
%reload_ext autotime
import pandas as pd
import requests
from pprint import pprint
import json
import torch
from PIL import Image
from transformers import MllamaForConditionalGeneration, AutoProcessor
from tqdm.auto import tqdm

pd.options.plotting.backend = "plotly"
pd.set_option("display.max_columns", None)
pd.set_option("display.max_colwidth", 100)
⌛ 1.44 µs (2024-12-11T12:18:00)
2024-12-11 12:18:03.892596: I tensorflow/core/util/port.cc:153] oneDNN custom operations are on. You may see slightly different numerical results due to floating-point round-off errors from different computation orders. To turn them off, set the environment variable `TF_ENABLE_ONEDNN_OPTS=0`.
2024-12-11 12:18:03.911171: E external/local_xla/xla/stream_executor/cuda/cuda_fft.cc:485] Unable to register cuFFT factory: Attempting to register factory for plugin cuFFT when one has already been registered
2024-12-11 12:18:03.931713: E external/local_xla/xla/stream_executor/cuda/cuda_dnn.cc:8454] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered
2024-12-11 12:18:03.938723: E external/local_xla/xla/stream_executor/cuda/cuda_blas.cc:1452] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered
2024-12-11 12:18:03.959172: I tensorflow/core/platform/cpu_feature_guard.cc:210] This TensorFlow binary is optimized to use available CPU instructions in performance-critical operations.
To enable the following instructions: AVX2 AVX512F AVX512_VNNI FMA, in other operations, rebuild TensorFlow with the appropriate compiler flags.
2024-12-11 12:18:04.791471: W tensorflow/compiler/tf2tensorrt/utils/py_utils.cc:38] TF-TRT Warning: Could not find TensorRT
In [2]:
df = pd.read_csv("results.csv").drop_duplicates(subset="panoid")
df
✔️ 20.5 ms (2024-12-11T12:18:05/2024-12-11T12:18:05)
Out[2]:
Index pid n time anxiousness latitude longitude geometry panoid panolat panolon panoyear panomonth
0 0 P20001 1 2023-04-25T02:51:42Z 0 -36.924795 174.738044 POINT (174.7380435 -36.92479483) QEpZV7bnO2mBfp0weMUKEg -36.924730 174.737826 2012 4
10 10 P20001 11 2023-04-24T00:42:25Z 0 -36.924837 174.737948 POINT (174.7379477 -36.92483659) pUw8PmVPYZBTW26mq2DAfw -36.924785 174.737728 2012 4
13 13 P20006 1 2023-06-03T02:45:55Z 3 -36.892203 174.740125 POINT (174.7401253 -36.89220256) omb98QNjTPWi0uUfMsmYeg -36.892621 174.739961 2024 5
14 15 P20009 2 2023-05-17T04:54:48Z 3 -36.923191 174.748620 POINT (174.7486203 -36.92319093) E7B5AV3DQ1rYWDClVRo8Zg -36.923194 174.748831 2024 5
17 19 P20009 6 2023-05-19T22:28:51Z 1 -36.923260 174.748655 POINT (174.748655 -36.92325959) KCTcsxYCIm41XdzkYEYUQw -36.923286 174.748840 2024 5
19 21 P20015 1 2023-05-17T07:34:00Z 5 -36.921603 174.747739 POINT (174.747739 -36.92160252) ESE0Slg2IO7Vf3QdBhETkg -36.921626 174.747253 2024 5
22 25 P20021 2 2023-06-04T02:33:49Z 6 -37.675727 175.209414 POINT (175.2094142 -37.67572725) NQi8eh4MBHprJGCpl9t1EQ -37.675820 175.209455 2023 11
23 26 P20021 3 2023-06-05T21:49:46Z 3 -36.894889 174.742775 POINT (174.7427751 -36.89488899) qgtMQGHZWUUIBCa8JgbBhA -36.895076 174.742734 2024 5
24 27 P20021 4 2023-06-06T02:29:11Z 5 -36.894854 174.742929 POINT (174.7429285 -36.89485419) T4yBf38jq472FmvtzEtI_w -36.895101 174.742848 2024 5
25 30 P20022 3 2023-04-25T06:42:09Z 1 -36.913380 174.731288 POINT (174.7312875 -36.91337995) do2cpZfBTwfxHkWnQkyL3A -36.913440 174.731310 2024 7
26 31 P20022 4 2023-04-25T22:31:15Z 6 -36.880662 174.707832 POINT (174.7078325 -36.88066162) vgFJoPSu-4SSfUKqz4BGgA -36.880785 174.707410 2024 6
27 33 P20022 6 2023-04-24T03:16:17Z 4 -36.852978 174.767267 POINT (174.7672665 -36.85297814) 0dU2CZ_GXhuhodJJbXVHSA -36.852731 174.767062 2017 2
28 34 P20027 1 2023-05-27T21:50:10Z 6 -36.892136 174.736943 POINT (174.7369429 -36.89213617) ody-NBwD6S0562GUtROqtg -36.891996 174.737012 2024 6
35 41 P20027 8 2023-05-30T21:17:36Z 2 -36.887537 174.736875 POINT (174.7368754 -36.88753691) Vy5UxGKwH8RxSoG2tFB94Q -36.887653 174.737623 2024 6
38 44 P20027 11 2023-06-01T05:24:54Z 3 -36.888974 174.735651 POINT (174.7356508 -36.88897381) IUzaXGEkzyOV-oEgTQaHhA -36.888938 174.735757 2023 1
39 45 P20027 12 2023-06-02T03:42:14Z 3 -36.887732 174.735789 POINT (174.7357892 -36.88773177) uPRCtL-OvwI3f_lXNBZtdw -36.888043 174.735442 2023 1
44 50 P20033 1 2023-05-03T08:19:03Z 0 -36.978477 174.830027 POINT (174.8300269 -36.9784771) BoCOn1VpFGrbXlyX3EKZ6g -36.978467 174.830275 2022 8
46 52 P20033 3 2023-05-04T02:17:26Z 0 -36.978365 174.830125 POINT (174.8301251 -36.97836488) tbfXbYFHITDw8p7vCFU3KA -36.978564 174.830225 2022 8
In [4]:
# Loading this model needs about 22.69GB of GPU memory
model_id = "meta-llama/Llama-3.2-11B-Vision-Instruct"

model = MllamaForConditionalGeneration.from_pretrained(
    model_id,
    torch_dtype=torch.bfloat16,
    device_map="auto",
)
processor = AutoProcessor.from_pretrained(model_id)
✔️ 16.7 s (2024-12-11T12:18:36/2024-12-11T12:18:53)
Downloading shards:   0%|          | 0/5 [00:00<?, ?it/s]
The model weights are not tied. Please use the `tie_weights` method before using the `infer_auto_device` function.
Loading checkpoint shards:   0%|          | 0/5 [00:00<?, ?it/s]
tokenizer_config.json:   0%|          | 0.00/55.8k [00:00<?, ?B/s]
chat_template.json:   0%|          | 0.00/5.09k [00:00<?, ?B/s]
In [7]:
for row in tqdm(df.head(10).itertuples(index=False)):
    panoid = row.panoid
    image = Image.open(f"panoramas/{panoid}.png")
    display(image)
    messages = [
        {
            "role": "user",
            "content": [
                {"type": "text", "text": """
                    This image is a panorama from Google Street View.
                    From the image, extract the following information, in JSON format:
                    green: Percentage of the image that is green space (e.g. parks, gardens, trees, grass etc.). A number from 0-100.
                    environment: Classify the nature of the environment in this image. Built up/green/residential/shops/cafes?. A string.
                    water: If you see any streams/ponds/rivers/ocean in the image, estimate the distance to the water in meters. A number. If there is no water, return 0.
                    obscured: Proportion of view obscured by buildings (how much of total line of sight is blocked by buildings in close proximity). A number from 0-100.
                    people: the number of people you see in the image
                    cars: the number of cars you see in the image
                    bikes: the number of bikes you see in the image

                    Do not include comments in your JSON response. Only respond with the JSON object. Make sure the JSON is valid.
                """},
                {"type": "image"},
            ]
        }
    ]
    input_text = processor.apply_chat_template(messages, add_generation_prompt=True)
    inputs = processor(
        image,
        input_text,
        add_special_tokens=False,
        return_tensors="pt"
    ).to(model.device)

    for retry in range(3):
        output = model.generate(**inputs, max_new_tokens=5000)
        result = processor.decode(output[0])
        result = result[result.rindex("<|end_header_id|>") + len("<|end_header_id|>"):].strip().replace("<|eot_id|>", "")
        print("Output:")
        try:
            result = json.loads(result)
            pprint(result)
            print("\n")
            break
        except json.JSONDecodeError:
            print(f"Unable to parse: {result}")
✔️ 26.2 s (2024-12-11T12:38:51/2024-12-11T12:39:18)
0it [00:00, ?it/s]
No description has been provided for this image
Output:
{'bikes': 0,
 'cars': 2,
 'environment': 'residential',
 'green': 40,
 'obscured': 60,
 'people': 1,
 'water': 0}


No description has been provided for this image
Output:
{'bikes': 0,
 'cars': 3,
 'environment': 'residential',
 'green': 30,
 'obscured': 0,
 'people': 0,
 'water': 0}


No description has been provided for this image
Output:
{'bikes': 0,
 'cars': 1,
 'environment': 'residential',
 'green': 90,
 'obscured': 10,
 'people': 0,
 'water': 0}


No description has been provided for this image
Output:
{'bikes': 0,
 'cars': 3,
 'environment': 'residential',
 'green': 55,
 'obscured': 45,
 'people': 0,
 'water': 0}


No description has been provided for this image
Output:
{'bikes': 0,
 'cars': 4,
 'environment': 'residential',
 'green': 50,
 'obscured': 50,
 'people': 0,
 'water': 0}


No description has been provided for this image
Output:
{'bikes': 0,
 'cars': 4,
 'environment': 'residential',
 'green': 70,
 'obscured': 10,
 'people': 0,
 'water': 0}


No description has been provided for this image
Output:
{'bikes': 0,
 'cars': 1,
 'environment': 'residential',
 'green': 70,
 'obscured': 0,
 'people': 0,
 'water': 0}


No description has been provided for this image
Output:
{'bikes': 0,
 'cars': 8,
 'environment': 'residential',
 'green': 75,
 'obscured': 20,
 'people': 0,
 'water': 0}


No description has been provided for this image
Output:
{'bikes': 0,
 'cars': 7,
 'environment': 'residential',
 'green': 50,
 'obscured': 30,
 'people': 0,
 'water': 0}


No description has been provided for this image
Output:
{'bikes': 0,
 'cars': 2,
 'environment': 'residential',
 'green': 67,
 'obscured': 49,
 'people': 0,
 'water': 0}